January 2018 - ghc-tickets

[GHC] #14732: -fdefer-type-holes breaks a correct program
by GHC 01 Feb '18

01 Feb '18

#14732: -fdefer-type-holes breaks a correct program -------------------------------------+------------------------------------- Reporter: MitchellSalad | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Here's a bug I discovered in `vector` that Ryan Scott identified as a regression from 7.10.3: https://github.com/haskell/vector/issues/200 Here is Ryan's minimal example: {{{#!hs {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE KindSignatures #-} {-# LANGUAGE MultiParamTypeClasses #-} module Bug where import Prelude hiding (zip, zipWith) zipWith :: (a -> b -> c) -> Bundle v a -> Bundle v b -> Bundle v c zipWith = undefined class GVector (v :: * -> *) a instance GVector Vector a data Bundle (v :: * -> *) a data Vector a class Unbox a stream :: GVector v a => v a -> Bundle v a {-# INLINE [1] stream #-} stream = undefined zip :: (Unbox a, Unbox b) => Vector a -> Vector b -> Vector (a, b) {-# INLINE [1] zip #-} zip = undefined {-# RULES "stream/zip [Vector.Unboxed]" forall as bs . stream (zip as bs) = zipWith (,) (stream as) (stream bs) #-}{#!hs }}} Output (8.2.2): {{{ Bug.hs:29:11: error: • Could not deduce (Unbox a) arising from a use of ‘zip’ from the context: GVector Vector (a, b) bound by the RULE "stream/zip [Vector.Unboxed]" at Bug.hs:(28,11)-(30,46) Possible fix: add (Unbox a) to the context of the RULE "stream/zip [Vector.Unboxed]" • In the first argument of ‘stream’, namely ‘(zip as bs)’ In the expression: stream (zip as bs) When checking the transformation rule "stream/zip [Vector.Unboxed]" | 29 | stream (zip as bs) = zipWith (,) (stream as) | ^^^^^^^^^ }}} Similar output in 8.0.1, 8.0.2, 8.2.1 -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14732> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #14723: GHC 8.4.1-alpha loops infinitely when typechecking
by GHC 01 Feb '18

01 Feb '18

#14723: GHC 8.4.1-alpha loops infinitely when typechecking -------------------------------------+------------------------------------- Reporter: RyanGlScott | Owner: (none) Type: bug | Status: new Priority: highest | Milestone: 8.4.1 Component: Compiler | Version: 8.4.1-alpha1 (Type checker) | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Compile-time Unknown/Multiple | performance bug Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- This issue prevents `jvm-streaming` from compiling with GHC 8.4.1-alpha2. Here is my best attempt at minimizing the issue: {{{#!hs {-# LANGUAGE DataKinds #-} {-# LANGUAGE FunctionalDependencies #-} {-# LANGUAGE KindSignatures #-} {-# LANGUAGE PolyKinds #-} {-# LANGUAGE RankNTypes #-} {-# LANGUAGE TypeFamilies #-} {-# LANGUAGE UndecidableInstances #-} module Bug () where import Data.Kind (Type) import Data.Proxy (Proxy(..)) import Data.String (fromString) import Data.Int (Int64) import GHC.Stack (HasCallStack) import GHC.TypeLits (Nat, Symbol) data JType = Iface Symbol data J (a :: JType) newIterator :: IO (J ('Iface "java.util.Iterator")) newIterator = do let tblPtr :: Int64 tblPtr = undefined iterator <- (loadJavaWrappers >> (((((((qqMarker (Proxy :: Proxy "{ return new java.util.Iterator() {\n @Override\n public native boolean hasNext();\n\n @Override\n public native Object next();\n\n @Override\n public void remove() {\n throw new UnsupportedOperationException();\n }\n\n private native v oid hsFinalize(long tblPtr);\n\n @Override\n public void finalize() {\n hsFinalize($tblPtr);\n }\n } ; }")) (Proxy :: Proxy "inline__method_0")) (Proxy :: Proxy "tblPtr")) (Proxy :: Proxy 106)) (tblPtr, ())) Proxy) (((callStatic (fromString "io.tweag.inlinejava.Inline__jvmstreaming022inplace_Language_Java_Streaming")) (fromString "inline__method_0")) [coerce tblPtr]))) undefined class Coercible (a :: Type) where type Ty a :: JType class Coercibles xs (tys :: k) | xs -> tys instance Coercibles () () instance (ty ~ Ty x, Coercible x, Coercibles xs tys) => Coercibles (x, xs) '(ty, tys) qqMarker :: forall -- k -- the kind variable shows up in Core (args_tys :: k) -- JType's of arguments tyres -- JType of result (input :: Symbol) -- input string of the quasiquoter (mname :: Symbol) -- name of the method to generate (antiqs :: Symbol) -- antiquoted variables as a comma-separated list (line :: Nat) -- line number of the quasiquotation args_tuple -- uncoerced argument types b. -- uncoerced result type (tyres ~ Ty b, Coercibles args_tuple args_tys, Coercible b, HasCallStack) => Proxy input -> Proxy mname -> Proxy antiqs -> Proxy line -> args_tuple -> Proxy args_tys -> IO b -> IO b qqMarker = undefined }}} With GHC 8.2.2, this is properly rejected by the typechecker: {{{ $ /opt/ghc/8.2.2/bin/ghc Bug.hs [1 of 1] Compiling Bug ( Bug.hs, Bug.o ) Bug.hs:27:12: error: Variable not in scope: loadJavaWrappers :: IO a0 | 27 | (loadJavaWrappers >> | ^^^^^^^^^^^^^^^^ Bug.hs:36:16: error: Variable not in scope: callStatic :: t0 -> t1 -> [a2] -> IO a1 | 36 | (((callStatic | ^^^^^^^^^^ Bug.hs:40:17: error: Variable not in scope: coerce :: Int64 -> a2 | 40 | [coerce tblPtr]))) | ^^^^^^ }}} But in GHC 8.4.1-alpha2, this simply hangs forever. To make things more interesting, if you pass `-ddump-tc-trace` when compiling, you'll get a panic: {{{ $ /opt/ghc/8.4.1/bin/ghc Bug.hs -ddump-tc-trace ... kcLHsQTyVars: cusk JType [] [] [] [] * [] [] [] [] kcTyClGroup: initial kinds [rB8 :-> ATcTyCon JType :: *, rB9 :-> APromotionErr RecDataConPE] txExtendKindEnv [rB8 :-> ATcTyCon JType :: *, rB9 :-> APromotionErr RecDataConPE] kcTyClDecl { JType env2 [] tcExtendBinderStack [] env2 [] tcExtendBinderStack [] lk1 Symbol tcTyVar2a Symbol * u_tys tclvl 1 * ~ TYPE t_a1qq[tau:1] arising from a type equality * ~ TYPE t_a1qq[tau:1] u_tys tclvl 1 'GHC.Types.LiftedRep ~ t_a1qq[tau:1] arising from a type equality * ~ TYPE t_a1qq[tau:1] u_tys tclvl 1 GHC.Types.RuntimeRep ~ GHC.Types.RuntimeRep arising from a kind equality arising from t_a1qq[tau:1] ~ 'GHC.Types.LiftedRep u_tys yields no coercion writeMetaTyVar t_a1qq[tau:1] :: GHC.Types.RuntimeRep := 'GHC.Types.LiftedRep u_tys yields no coercion u_tys yields no coercion checkExpectedKind * TYPE t_a1qq[tau:1] <*>_N kcLHsQTyVars: not-cuskghc: panic! (the 'impossible' happened) (GHC version 8.4.0.20180118 for x86_64-unknown-linux): kcConDecl }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14723> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 9

[GHC] #14715: GHC 8.4.1-alpha regression with PartialTypeSignatures
by GHC 01 Feb '18

01 Feb '18

#14715: GHC 8.4.1-alpha regression with PartialTypeSignatures -------------------------------------+------------------------------------- Reporter: RyanGlScott | Owner: (none) Type: bug | Status: new Priority: highest | Milestone: 8.4.1 Component: Compiler | Version: 8.4.1-alpha1 (Type checker) | Keywords: | Operating System: Unknown/Multiple PartialTypeSignatures | Architecture: | Type of failure: GHC rejects Unknown/Multiple | valid program Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- This bug prevents `lol-apps`' tests and benchmarks from building with GHC 8.4.1-alpha2. This is as much as I'm able to minimize the issue: {{{#!hs {-# LANGUAGE FlexibleContexts #-} {-# LANGUAGE MultiParamTypeClasses #-} {-# LANGUAGE PartialTypeSignatures #-} {-# LANGUAGE ScopedTypeVariables #-} {-# LANGUAGE TypeFamilies #-} {-# OPTIONS_GHC -Wno-partial-type-signatures #-} module Bug (bench_mulPublic) where data Cyc r data CT zp r'q class Reduce a b type family LiftOf b bench_mulPublic :: forall z zp zq . (z ~ LiftOf zq, _) => Cyc zp -> Cyc z -> IO (zp,zq) bench_mulPublic pt sk = do ct :: CT zp (Cyc zq) <- encrypt sk pt undefined ct encrypt :: forall z zp zq. Reduce z zq => Cyc z -> Cyc zp -> IO (CT zp (Cyc zq)) encrypt = undefined }}} On GHC 8.2.2, this compiles without issue. But on GHC 8.4.1-alpha2, this errors with: {{{ $ /opt/ghc/8.4.1/bin/ghc Bug.hs [1 of 1] Compiling Bug ( Bug.hs, Bug.o ) Bug.hs:15:1: error: • Could not deduce (Reduce fsk0 zq) from the context: (z ~ LiftOf zq, Reduce fsk zq) bound by the inferred type for ‘bench_mulPublic’: forall z zp zq fsk. (z ~ LiftOf zq, Reduce fsk zq) => Cyc zp -> Cyc z -> IO (zp, zq) at Bug.hs:(15,1)-(17,14) The type variable ‘fsk0’ is ambiguous • In the ambiguity check for the inferred type for ‘bench_mulPublic’ To defer the ambiguity check to use sites, enable AllowAmbiguousTypes When checking the inferred type bench_mulPublic :: forall z zp zq fsk. (z ~ LiftOf zq, Reduce fsk zq) => Cyc zp -> Cyc z -> IO (zp, zq) | 15 | bench_mulPublic pt sk = do | ^^^^^^^^^^^^^^^^^^^^^^^^^^... }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14715> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #14725: memory leak: forkOS not releasing thread local `Task` struct upon end of thread
by GHC 01 Feb '18

01 Feb '18

#14725: memory leak: forkOS not releasing thread local `Task` struct upon end of thread ------------------------------------------+--------------------------- Reporter: RobertZabel | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Runtime System | Version: 8.2.2 Keywords: forkOS memory leak | Operating System: POSIX Architecture: Unknown/Multiple | Type of failure: Other Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: ------------------------------------------+--------------------------- I came to notice a memory leak in the runtime system while letting warp use `forkOS` and conducting a load test.\\ A minimal snippet to reproduce: {{{#!hs module Main where import Control.Concurrent main :: IO () main = replicateM_ 10000000 $ forkOS $ return () }}} This patch will clean up the thread local `Task` struct just before threads created by forkOS terminate. {{{ diff --git a/rts/posix/OSThreads.c b/rts/posix/OSThreads.c index c9adf4e..bd4e75d 100644 --- a/rts/posix/OSThreads.c +++ b/rts/posix/OSThreads.c @@ -223,6 +223,7 @@ forkOS_createThreadWrapper ( void * entry ) cap = rts_lock(); rts_evalStableIO(&cap, (HsStablePtr) entry, NULL); rts_unlock(cap); + rts_done(); return NULL; } }}} I think win32 platforms need an equivalent patch, but unfortunately I cannot verify that. \\ Hence I'm only targeting posix here. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14725> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #14702: Runtime is keeping hold to too much memory
by GHC 01 Feb '18

01 Feb '18

#14702: Runtime is keeping hold to too much memory -------------------------------------+------------------------------------- Reporter: blaze | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Runtime | Version: 8.2.2 System | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Other Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Currently runtime only releases memory to OS when it has more than 4x of the amount used. This leads to heavy memory overuse when process is run using many processes and large nurseries. For example, if process is run on 16 CPUs with each having 32MB nursery, it will use total 512MB allocation area. This will push "release boundary" to the 2GB. If process is ever to allocate that much memory for a short term use, it will never be given back to OS and effectively wasted. Nursery and other non-growing data areas should be excluded from this multiplier. Runtime should only keep an extra memory proportional to heap allocation. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14702> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #10855: GHC rejects code that Haskell 2010 report accepts
by GHC 01 Feb '18

01 Feb '18

#10855: GHC rejects code that Haskell 2010 report accepts -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- According to [https://www.haskell.org/onlinereport/haskell2010/haskellch3.html#x8-220003 The Haskell 2010 Report], the following two modules should compile. But they don't, failing with parse errors: {{{ {-# LANGUAGE Haskell2010 #-} module Bug where bool = - case b of False -> 0; True -> (-1) }}} {{{ {-# LANGUAGE Haskell2010 #-} instance Num (IO a) where negate = id main = - do putStrLn "hi!" }}} Note the unary `-` signs. I'm not sure whether we should complicate the parser to fix these, but we should document the compliance-failure in the manual. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10855> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #10074: Implement the 'Improved LLVM Backend' proposal
by GHC 01 Feb '18

01 Feb '18

#10074: Implement the 'Improved LLVM Backend' proposal -------------------------------------+------------------------------------- Reporter: | Owner: thoughtpolice thoughtpolice | Status: new Type: task | Milestone: 7.12.1 Priority: high | Version: Component: Compiler | Operating System: Unknown/Multiple (LLVM) | Type of failure: None/Unknown Keywords: llvm, | Blocked By: codegen | Related Tickets: Architecture: | Unknown/Multiple | Test Case: | Blocking: | Differential Revisions: Phab:D530 | -------------------------------------+------------------------------------- This is a meta ticket designed to reflect the current implementation status of the 'Improved LLVM Backend' proposal, documented here: https://ghc.haskell.org/trac/ghc/wiki/ImprovedLLVMBackend -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10074> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 34

[GHC] #14695: ghc --interactive error: "gcc: error: libgmp.so: No such file or directory"
by GHC 01 Feb '18

01 Feb '18

#14695: ghc --interactive error: "gcc: error: libgmp.so: No such file or directory" -------------------------------------+------------------------------------- Reporter: osa1 | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- I get this error when I run hello world with `ghci` or `runghc`: {{{ $ ghc-stage2 --interactive Hello.hs GHCi, version 8.5.20180117: http://www.haskell.org/ghc/ :? for help gcc: error: libgmp.so: No such file or directory gcc: fatal error: no input files compilation terminated. `gcc' failed in phase `gcc'. (Exit code: 1) }}} It works with 8.2.2's `runghc` and `ghci` so it seems like a regression to me. This also causes validate failure on my laptop, because at one point validate script runs `runghc`. OS: Xubuntu 16.04.2 `libgmp.so` location: {{{ ➜ ~ locate libgmp.so /usr/lib/x86_64-linux-gnu/libgmp.so /usr/lib/x86_64-linux-gnu/libgmp.so.10 /usr/lib/x86_64-linux-gnu/libgmp.so.10.3.0 /usr/lib/x86_64-linux-gnu/openssl-1.0.0/engines/libgmp.so }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14695> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

[GHC] #12245: Deriving Data at higher kinds
by GHC 01 Feb '18

01 Feb '18

#12245: Deriving Data at higher kinds -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Here's what [https://mail.haskell.org/pipermail/generics/2016-June/000564.html Lennart Spitzner wanted to do]: {{{ > {-# LANGUAGE StandaloneDeriving #-} > {-# LANGUAGE DeriveDataTypeable #-} > {-# LANGUAGE FlexibleInstances #-} > > import Data.Data ( Data ) > > data Foo f = Foo (f Bool) (f Int) > > deriving instance Data (Foo []) > deriving instance Data (Foo Maybe) }}} Of course you can't derive `Data` for `Foo` because we don't know what `f` is, so Lennart is making multiple instances, one for each instance of `f`. It's a bit clumsy. What we would really like is {{{ deriving instance (forall a. Data => Data (f a)) => Data (Foo f) }}} but we don't have higher order instances yet! So Lennart is manually making two instances. This should work, but he gets {{{ > Main.hs: line 45, column 1: > Multiple declarations of ‘$cr2C’ > Declared at: Main.hs:44:1 > Main.hs:45:1 > Main.hs: line 45, column 1: > Multiple declarations of ‘$tr2D’ > Declared at: Main.hs:44:1 > Main.hs:45:1 }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12245> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

[GHC] #14687: Investigate differences in Int-In/Outlining
by GHC 31 Jan '18

31 Jan '18

#14687: Investigate differences in Int-In/Outlining -------------------------------------+------------------------------------- Reporter: AndreasK | Owner: (none) Type: task | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- In one case the numbers get floated out and in the others they stay in the alternative. This results in one being compiled as a CAF while the other gets recreated on the heap. {{{ {-# LANGUAGE MagicHash, BangPatterns, ScopedTypeVariables #-} import GHC.Prim import GHC.Exts func :: Int# -> Int# -> Int func 1# 1# = 10 func 1# 2# = 20 foo :: Int# -> Int# -> Int foo 1# 1# = 30 foo 2# 1# = 40 }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14687> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 2